I have a data set where the monthly sales of TMT bars and various other explanatory variables are present from April 2014-March 2018. I need to predict the monthly sales of the coming/next month. Since I do not know the values of few explanatory variables beforehand, hence I have taken the approach, where I take data until November 2017 as train and prepared the model. Then using the November 2017's(single month's) data to test the model. Similarly, I have prepared the model on data till December 2017, and then tested it on only December 2017's data and so on.Hence the final model will be prepared data till March 2018 and it will be tested on March 2018's data point. Is this the correct approach? What other approaches should I consider? Thanks in advance.
1 Answer
$\begingroup$
$\endgroup$
Yes you can be predicting the input series and then using the predictions to predict the output series (Y) . Care should be taken to incorporate the uncertainty in the predictions of the input series (the x'S) INTO the uncertainty of the final forecast of Y.
